license: mit
library_name: transformers
pipeline_tag: text-generation
🌐 WebThinker-R1-14B
WebThinker: Empowering Large Reasoning Models with Deep Research Capability
Large reasoning models (LRMs), such as OpenAI-o1 and DeepSeek-R1, demonstrate impressive long-horizon reasoning capabilities. However, their reliance on static internal knowledge limits their performance on complex, knowledge-intensive tasks and hinders their ability to produce comprehensive research reports requiring synthesis of diverse web information. To address this, we propose WebThinker, a deep research agent that empowers LRMs to autonomously search the web, navigate web pages, and draft research reports during the reasoning process. WebThinker integrates a Deep Web Explorer module, enabling LRMs to dynamically search, navigate, and extract information from the web when encountering knowledge gaps. It also employs an Autonomous Think-Search-and-Draft strategy, allowing the model to seamlessly interleave reasoning, information gathering, and report writing in real time. To further enhance research tool utilization, we introduce an RL-based training strategy via iterative online Direct Preference Optimization (DPO). Extensive experiments on complex reasoning benchmarks (GPQA, GAIA, WebWalkerQA, HLE) and scientific report generation tasks (Glaive) demonstrate that WebThinker significantly outperforms existing methods and strong proprietary systems. Our approach enhances LRM reliability and applicability in complex scenarios, paving the way for more capable and versatile deep research systems.
Overview
WebThinker-R1-14B is part of the WebThinker series that enables large reasoning models to autonomously search, explore web pages, and draft research reports within their thinking process. This 14B parameter model provides deep research capabilities through:
- Deep Web Exploration: Enables autonomous web searches and page navigation by clicking interactive elements to extract relevant information while maintaining reasoning coherence
- Autonomous Think-Search-and-Draft: Integrates real-time knowledge seeking with report generation, allowing the model to draft sections as information is gathered
- RL-based Training: Leverages iterative online DPO training with preference pairs constructed from reasoning trajectories to optimize end-to-end performance
Related Models
Usage
This model can be used for:
- Complex problem solving requiring external knowledge
- Scientific research report generation
- Open-ended reasoning tasks
Citation
@article{Li2025WebThinker,
author = {Xiaoxi Li and
Jiajie Jin and
Guanting Dong and
Hongjin Qian and
Yutao Zhu and
Yongkang Wu and
Ji{-}Rong Wen and
Zhicheng Dou},
title = {WebThinker: Empowering Large Reasoning Models with Deep Research Capability},
journal = {CoRR},
volume = {abs/2504.21776},
year = {2025},
url = {https://arxiv.org/abs/2504.21776},
doi = {10.48550/ARXIV.2504.21776},
eprinttype = {arXiv},
eprint = {2504.21776}
}
License
This model is released under the MIT License.
Contact
For any questions or feedback, please reach out to us at xiaoxi_li@ruc.edu.cn.